DeepSeek V4 Modell updates?

[Thomas Leclaire (tleclaire)]

Hi,

When can we expect the latest DeepSeek V4 Flash and Pro Model updates to be available in Foundry?

[Klay (Shyboy0499)]

Good news — they're already in the catalog. DeepSeek-V4-Flash and DeepSeek-V4-Pro are available in the Microsoft Foundry model catalog under the "Direct from Azure" collection (catalog versions dated 2026-04-23 as of now), deployable as serverless/managed endpoints without importing the weights yourself.

If you don't see them, the usual cause is regional rollout or UI/cache lag rather than the models not existing.

How new versions land, and where to track them

1. The model catalog is the source of truth. In Foundry, go to Model catalog → filter to "Direct from Azure" → search DeepSeek. The model card shows the current version and status (Preview vs GA).
2. "What's New" in the Foundry docs and the Azure Updates feed (azure.microsoft.com/updates) are where new catalog entries and GA announcements get posted.
3. Region availability pages list which regions serve each DeepSeek version — availability rolls out region-by-region, so one region can lag another.
4. The MicrosoftDocs/azure-ai-docs repo on GitHub tracks catalog changes in its "what's new" change logs, if you want the raw feed.

Things to check if a model looks missing

• Confirm your workspace region is listed on the availability page for that model.
• Filter to "Direct from Azure" — DeepSeek appears under that collection, not the open-source one.
• "Preview" labels can persist in the UI even after GA due to rollout/caching; a hard refresh or re-login usually clears it.

On timing

Foundry doesn't publish exact release dates ahead of time — the practical cadence is "the moment a version appears in the catalog." If you need a specific version on a deadline, watch the catalog for the version bump, and open a support request for a concrete ETA; that's the only channel with one.